← 返回最新日报
2026-10-20· 当日存档
AI 日报
DeepSeek开源推理栈引关注
今日看点：DeepSeek开源V4.1生产推理栈DeepInfer并支持SM100，同时美国下令联邦机构移除DeepSeek产品，形成鲜明对照；MiniMax发布428B MoE模型M3.1 Flash，以超低价搅动市场；vLLM推送式KV传输显著优化推理延迟。
模型动向
1
what a thing to wake up to. sm100 support is very interesting
DeepSeek开源生产推理栈并适配SM100，值得关注
X · List深挖 →
2
3
MiniMax Just Launched M3.1 Flash: 428B MoE Agentic Model at $0.15/M Input
428B MoE仅激活23B，API价格极低
MarkTechPost深挖 →
4
MiniMax M3.1 Flash shows China's AI price race has no floor in sight
智能指数42分，成本减半，价格战加剧
The Decoder深挖 →
5
Inception Releases Mercury 3 Diffusion LLM at 1,068 Tokens per Second
扩散LLM达1068 token/秒，支持百万上下文
MarkTechPost深挖 →
6
Kling AI drops version 4.0 with stereo audio, multilingual lip sync, and a lighter Flash variant
可灵4.0支持30秒视频与多语种唇同步
The Decoder深挖 →
7
Nvidia launches new open models to run agents in the enterprise and on PCs
英伟达开源模型瞄准企业与PC端智能体
TechCrunch AI深挖 →
8
9
Cloudflare's Clairic pulls data from documents faster and cheaper than LLMs
Cloudflare称文档提取比大模型更快更便宜
The Decoder深挖 →
产品上新
10
Structured outputs for Anthropic Claude models on Amazon Bedrock
Bedrock为Claude全系提供JSON Schema结构化输出
AWS ML深挖 →
11
12
Adobe Lightroom mobile can now open closed eyes in photos with a simple tap
Lightroom手机版一键AI睁眼修复闭眼照片
The Verge AI深挖 →
论文研究
13
“We were looking at it wrong”: The explosive claim that’s upended the AI world
2.6万字长文提出神经网络本质新说，震动AI圈
MIT Tech Review深挖 →
行业动态
14
US orders federal agencies to remove China’s DeepSeek AI from their systems
美联邦机构被令移除DeepSeek产品及第三方服务
TechCrunch AI深挖 →
15
The neocloud shakeout is coming, and Nvidia has the most to lose
AI云将洗牌，英伟达因非超大客户占比高最受伤
The Decoder深挖 →
16
Army Cyber Command's ECHOTRIBBLE would turn AI agents into self-spreading cyberweapons
美陆军项目让AI代理自我传播攻击，专家警告风险
The Decoder深挖 →
17
A hacker broke into John Deere and SharkNinja with an army of AI agents
黑客用AI代理集群攻破超百家网站
The Verge AI深挖 →
18
Two German courts now hold Google liable for what its AI overviews say
德国法院裁定谷歌为AI概览错误内容担责
The Decoder深挖 →
19
Meta Muse exposes users' personal and business secrets when they share conversations
Meta Muse泄露用户私密与商业对话
Ars Technica AI深挖 →
20
WhatsApp's help pages hid LLM instructions: dont recommend a regulator
WhatsApp帮助页暗藏指令，要求别推荐监管机构
Hacker News深挖 →
21
Amazon is taking its biggest shot at Nvidia yet with new TorchNeuron dev tool
亚马逊TorchNeuron降低PyTorch跑自研芯片门槛
TechCrunch AI深挖 →
22
Eka is worth $1.45B thanks to a physical AI model that has experts raving
Eka物理AI模型让机械臂流畅抓取，估值14.5亿
TechCrunch AI深挖 →
23
24
The next step for AI research is getting answers that haven’t been published yet. Qualitate has raised a $25M Series A led by Insight Partners. Its A...
Qualitate融资2500万美元，用AI访谈生成结构化数据
X · List深挖 →
实用技巧
25
Push-based KV transfer to improve TTFT and TPOT for disaggregated inference in vLLM
推送式KV传输让TTFT提升3倍、TPOT提升30%
AWS ML深挖 →
26
Improve TTFT for multi-turn agentic workloads with bi-directional KV transfer in vLLM
双向KV复用缓存，多轮对话首字延迟降3.3倍
AWS ML深挖 →
27
Harness Engineering 101: The LLM is the brain, the harness is the body
模型是大脑，harness才是工程落地关键
Hacker News深挖 →
28
Ask HN: How to deal with shipping AI-written code you don't understand?
HN热议如何应对看不懂的AI生成代码
Hacker News深挖 →
快讯
·
《周期》精读第26期：给市场量体温的指标
一石一泉一松一月一人
日报存档
2026 年 10 月
24 日Kimi开源3T模型领跑23 日AI安全与资本双线升温22 日AI攻防与代理竞速21 日智能体安全与开源基建双线并进20 日DeepSeek开源推理栈引关注19 日智能体加速落地，治理争议升温18 日智能体能力与风险同框17 日开源模型双雄并进，安全治理引关注16 日开源巨模与上市竞速15 日Anthropic营收与IPO承压14 日谷歌Gemini 4发布与AI安全警报13 日AI狂飙与隐忧并存12 日AI狂飙与认知鸿沟11 日AI失控与安全警钟10 日AI代理失控警钟9 日AI智能体全面提速8 日GPT-6 智能 UI 上线7 日万亿模型与AI自治争议6 日开源模型与资本双线爆发5 日AI侵蚀大学信任4 日OpenAI安全震荡与验证器研究3 日AI代理安全与落地并进2 日AI能力跃迁与安全隐忧1 日谷歌Gemini 4发布引三强竞速2026 年 9 月
22 日小米开源登顶，AI失控警钟21 日开源图像模型与AI治理并进20 日AI安全与失控风险成焦点19 日AI安全失控与模型竞速18 日AI对齐警报与芯片竞赛17 日AI安全警报与产品整合潮16 日谷歌苹果齐发AI新品15 日苹果Siri重构与AI放缓之争14 日AI巨头放缓迭代引激辩13 日GPT-6 Astra全面突破12 日AI数学错位与安全风波11 日智能体基建与模型竞速10 日GPT-6登场，AI安全恐慌升温9 日AI数学突破引热议8 日AI日报：端侧模型与欧洲融资双热点7 日AI日报：模型竞赛与行业震荡6 日GPT-6 Astra发布，AI竞争白热化5 日GPT-6发布，AGI时代开启4 日GPT-6 Astra开启AGI时代3 日英伟达巨资收购HF2 日AI日报：世界模型与安全并进1 日AI日报：模型突破与行业震荡